Before AI Can Understand Human Behaviour, It Must Understand Human Systems

Abstract

Artificial intelligence has achieved remarkable progress in language processing, legal reasoning and medical decision support. Yet when applied to human relationships, most current approaches continue to analyze conversations, behavioural descriptions or psychological symptoms in isolation. This article argues that the fundamental limitation is not computational power, but representation.

Human relationships emerge from adaptive systems composed of interacting psychological functions operating under uncertainty. Before artificial intelligence can meaningfully reason about human behaviour, those systems must first be explicitly modelled.

Rather than replacing psychology, systems theory may provide the structural language needed to represent how attachment, memory, identity, defensive mechanisms, desire, validation and conscious decision-making interact over time.

This article proposes the conceptual foundations for such a model and outlines its potential applications in artificial intelligence, psychology, relational analysis and interdisciplinary research.

1. The Current Limits of Relational AI

Artificial intelligence has transformed the way information is processed. Large language models are now capable of generating legal analyses, assisting medical diagnosis, producing software and participating in increasingly sophisticated forms of reasoning. Their success has been built upon an extraordinary ability to process vast amounts of structured and unstructured information while identifying linguistic and statistical patterns that would be impossible for humans to analyse at the same scale.

When artificial intelligence is applied to human relationships, however, the underlying methodology remains remarkably similar. Most existing systems continue to analyse conversations, emails, text messages or behavioural descriptions as their primary source of information. Users present fragments of dialogue or describe interpersonal events, expecting the model to infer emotions, motivations or even predict future behaviour from the language itself.

This approach undoubtedly provides valuable insights. Language reflects emotional states, cognitive processes and relational dynamics. Patterns of communication may reveal attachment strategies, defensive reactions or significant behavioural changes. Nevertheless, language is not the system being analysed. It is only one observable expression of a considerably more complex internal architecture.

The distinction is familiar in many scientific disciplines. A physician does not diagnose a patient solely from a symptom. A lawyer does not construct legal reasoning from a single sentence extracted from a contract. An aerospace engineer does not evaluate an entire spacecraft by observing only one sensor. In every case, individual observations acquire meaning because they are interpreted within a structured model of the underlying system.

Human relationships should not necessarily be approached differently. Conversations, behaviours and psychological observations represent valuable evidence, but they remain individual manifestations of a deeper relational system. Without an explicit representation of that system, artificial intelligence is forced to reason primarily from observable outputs rather than from the mechanisms that generate them.

The central argument of this article is therefore methodological rather than technological. The principal limitation of relational artificial intelligence is unlikely to be computational capability or the quality of language models themselves. Instead, it lies in the absence of a systems model capable of representing how the psychological functions underlying human behaviour interact, adapt and evolve over time. Before artificial intelligence can meaningfully understand human relationships, it must first understand the systems from which those relationships emerge.

2. Human Behaviour Is Not Language

One of the most common assumptions in current relational artificial intelligence is that language provides sufficient information to explain human behaviour. Conversations, messages and written narratives undoubtedly contain valuable evidence, yet they remain observations rather than explanations. They describe what people communicate, but not necessarily the internal system that produced those communications.

Human behaviour does not emerge directly from language. It emerges from a continuously evolving interaction between conscious thought, subconscious processes, memory, emotional regulation, attachment, defence mechanisms and environmental influences. Language is simply one of the observable outputs generated by this much broader internal architecture.

This distinction is fundamental. Two individuals may produce almost identical conversations while operating under entirely different psychological conditions. A brief message may express affection, avoidance, uncertainty or emotional self-protection depending on the internal state of the relational system at that particular moment. Analysing language without representing the system behind it risks confusing observable behaviour with the mechanisms responsible for generating it.

The same principle applies across many scientific disciplines. Engineers distinguish between sensor readings and the machine itself. Physicians distinguish between symptoms and the biological processes producing them. Likewise, relational analysis should distinguish between communication and the psychological system from which communication emerges. Conversations are evidence, but they are not the system itself.

This distinction also explains why human relationships often appear contradictory when analysed exclusively through language. Individuals may consciously express intentions that differ from their subconscious responses. Emotional attachment, defensive adaptation, identity and personal history may simultaneously influence behaviour, producing actions that appear inconsistent when interpreted only at the level of verbal communication.

If artificial intelligence is expected to reason meaningfully about human relationships, it cannot simply become better at analysing conversations. It must first represent the internal architecture that gives those conversations meaning. Human behaviour is not language. Language is one observable manifestation of a dynamic system whose structure remains largely unmodelled.

3. Psychology Describes Functions, Systems Theory Describes Interactions

Modern psychology has made remarkable progress in identifying the internal mechanisms that influence human behaviour. Attachment theory, cognitive psychology, behavioural science and neuroscience have described numerous psychological functions that contribute to emotional regulation, interpersonal relationships and decision-making. Concepts such as attachment, identity, defence mechanisms, emotional memory and motivation are now well established within contemporary psychological research.

These contributions have significantly improved our understanding of human behaviour. They explain why individuals develop different attachment styles, why defensive responses emerge under emotional stress and how previous experiences continue to influence present relationships. In many respects, psychology has successfully identified many of the functional components that participate in human relational systems.

Yet identifying individual functions is not necessarily equivalent to modelling the system in which those functions operate. A systems perspective shifts the focus from the components themselves to the relationships that exist between them. The central question is no longer simply what functions exist, but rather how those functions interact, compete, cooperate and adapt over time.

This distinction may prove particularly important for artificial intelligence. Language models can recognise psychological concepts when they appear in text, but understanding a relational system requires more than identifying isolated functions. It requires representing how conscious reasoning, subconscious processes, attachment, emotional regulation, memory and defensive mechanisms continuously influence one another while responding to changing environmental conditions.

From this perspective, systems theory does not seek to replace psychology. Instead, it offers a complementary structural language capable of describing the interactions between psychological functions. Psychology explains the existence of many individual processes. Systems theory attempts to explain the architecture that allows those processes to generate coherent — and sometimes apparently contradictory — patterns of human behaviour.

This distinction may ultimately represent the missing step between contemporary psychology and future relational artificial intelligence. Before human behaviour can be modelled computationally, the interactions between its underlying psychological functions must first be represented within a coherent systems framework. Only then can artificial intelligence move beyond analysing isolated observations toward reasoning about the dynamic structure from which those observations emerge.

4. Toward a Systems Model of Human Behaviour

If human behaviour is understood as the observable expression of an internal system, the next question becomes inevitable: what exactly should that system represent? A systems model cannot simply reproduce conversations or behavioural descriptions. Its purpose is to represent the internal architecture from which those observable behaviours emerge.

Unlike traditional language analysis, a systems approach begins by identifying the principal functional components participating in human behaviour. These components should not be interpreted as isolated psychological traits but as dynamic functions capable of interacting, activating, adapting and occasionally competing with one another. Human behaviour therefore becomes the emergent result of multiple simultaneous processes rather than the direct consequence of a single emotional or cognitive state.

A preliminary systems model might therefore include conscious reasoning, subconscious processes, attachment, emotional memory, identity, defensive adaptation, motivational functions, environmental influences and temporal evolution. None of these elements operates independently. Every function continuously influences the others, generating an adaptive system whose behaviour changes as both internal and external conditions evolve.

One important consequence of this perspective is that relationships themselves become systems rather than static emotional states. Attachment, desire, trust, validation, cooperation or emotional distance should not necessarily be viewed as isolated phenomena. Instead, they may represent temporary configurations produced by the interaction of multiple underlying psychological functions operating simultaneously within the same relational architecture.

Such a model does not seek to reduce human experience to mechanical processes. Like every scientific model, it is an abstraction designed to simplify reality sufficiently to make systematic analysis possible. The objective is not to eliminate complexity but to organise it into a coherent structural framework capable of supporting explanation, comparison and computational reasoning.

Developing such a framework represents a considerable interdisciplinary challenge. Psychology provides many of the individual functional components. Systems theory offers the language to describe their interactions. Artificial intelligence may eventually provide the computational tools capable of reasoning over those interactions. Bringing these disciplines together may represent the first step toward a genuine systems model of human behaviour.

5. Uncertainty as a Structural Property

One of the most common objections to modelling human behaviour is that people are inherently unpredictable. Human decisions are influenced by emotions, experience, personal history, social context and countless external variables. Unlike mechanical systems, human beings rarely respond identically under apparently identical conditions. At first sight, this uncertainty may appear incompatible with any attempt to construct a systems model.

The opposite may be true. Uncertainty should not necessarily be viewed as a limitation of the model, but as one of its fundamental properties. Many complex systems in engineering, economics and biology cannot be described through deterministic rules alone. They are understood through probabilistic models capable of representing multiple possible states and the conditions under which transitions between those states become more or less likely.

Human behaviour may be approached in a similar manner. The objective is not to predict individual decisions with absolute certainty, but to understand how different psychological functions interact to increase or reduce the probability of particular behaviours. Artificial intelligence would therefore reason not in terms of certainties, but in terms of evolving system states and adaptive probabilities.

This perspective also offers an alternative interpretation of many psychological processes. Defensive mechanisms, for example, may be understood not merely as behavioural responses but as regulatory functions intended to reduce perceived uncertainty and restore internal equilibrium. Attachment dynamics, emotional distancing or relational withdrawal may represent adaptive attempts to stabilise the system under changing internal or external conditions.

A systems approach therefore shifts the focus away from isolated behaviours and towards the conditions that make those behaviours more or less probable. Human systems remain adaptive, dynamic and partially unpredictable, yet they are not necessarily random. Their uncertainty forms part of their architecture rather than representing a failure of explanation.

Recognising uncertainty as a structural property may ultimately prove essential for future relational artificial intelligence. Instead of attempting to predict exactly what an individual will do, AI may become capable of modelling how the internal state of a human system evolves and which behavioural transitions become increasingly probable under specific conditions.

6. From Pattern Recognition to System Prediction

Contemporary artificial intelligence has demonstrated extraordinary capabilities in recognising patterns within language. Modern models can identify emotional tone, infer probable intentions, detect inconsistencies and generate sophisticated interpretations from relatively limited textual information. These abilities explain much of the recent success of large language models across numerous professional disciplines.

Relational analysis, however, presents a fundamentally different challenge. Human relationships do not evolve through language alone. Conversations represent only one observable layer of a much broader adaptive system whose internal state continuously changes through the interaction of psychological functions, memory, environmental influences and individual decision-making.

A future relational AI may therefore require a different methodological approach. Rather than interpreting isolated conversations, it could begin by constructing an explicit representation of the underlying human system. Language would remain an essential source of information, but it would no longer constitute the primary object of analysis. Instead, conversations would become observable evidence from which the internal state of the system could be progressively inferred.

Such an approach would also transform the nature of prediction. The objective would no longer be to determine with certainty what an individual will do next. Human systems remain adaptive and inherently uncertain. Instead, artificial intelligence could estimate which systemic transitions become increasingly probable given the current configuration of the model and the interactions between its psychological functions.

This distinction may appear subtle, yet it fundamentally changes the role of artificial intelligence. Instead of generating interpretations based primarily upon linguistic patterns, AI would begin reasoning about evolving system states. Behaviour would no longer be viewed as an isolated event requiring explanation, but as the observable consequence of a dynamic architecture continuously adapting to internal and external conditions.

Whether such systems can eventually be constructed remains an open question. Nevertheless, the transition from pattern recognition towards system prediction may represent one of the most significant conceptual challenges facing future artificial intelligence. Larger language models alone may not be sufficient. More comprehensive representations of human systems may ultimately prove equally important.

7. Beyond Relationships: A General Framework

Although this discussion has focused primarily on human relationships, the underlying principles extend far beyond interpersonal attachment. If human behaviour can be represented as the interaction of adaptive psychological functions operating within a dynamic system, the same conceptual framework may prove relevant across numerous disciplines concerned with human decision-making and organisational behaviour.

Psychology represents perhaps the most immediate application. A systems language may complement existing psychological theories by providing a structured representation of how multiple functions interact over time rather than analysing each process in relative isolation. The objective is not to replace established psychological models but to offer a common framework capable of integrating them into a coherent dynamic architecture.

Artificial intelligence constitutes a second natural application. Current language models demonstrate remarkable abilities in analysing communication, yet future systems may increasingly require explicit representations of human systems capable of supporting probabilistic reasoning, adaptive modelling and relational prediction under conditions of uncertainty.

The same principles may also prove valuable within law, organisational analysis and sociology. Legal disputes frequently involve evolving human systems rather than isolated events. Organisations continuously adapt through interacting decision-making processes. Social institutions emerge from collective behavioural dynamics extending across multiple individuals and environments. In each case, understanding the interactions between system components may become as important as analysing the individual components themselves.

This broader perspective illustrates that systems theory should not be understood as a competing discipline, but as a common structural language capable of connecting knowledge developed independently across psychology, artificial intelligence, engineering and the social sciences. Different disciplines may continue studying different aspects of human behaviour while sharing a common framework for representing the interactions between them.

Whether such a unified systems language can ultimately be developed remains an open research question. Nevertheless, the growing convergence between artificial intelligence, cognitive science and systems thinking suggests that the search for more comprehensive models of human behaviour has only just begun.

8. Conclusion

Artificial intelligence has demonstrated extraordinary capabilities in processing language, recognising patterns and assisting human decision-making across numerous professional disciplines. Nevertheless, when applied to human relationships, most current approaches continue to analyse conversations, behavioural descriptions and isolated observations rather than the underlying systems from which those observations emerge.

This article has argued that the principal challenge may not be computational but conceptual. Before artificial intelligence can reason meaningfully about human behaviour, the object of analysis itself must first be represented. Human relationships do not arise from language alone. They emerge from adaptive systems composed of interacting psychological functions continuously evolving under conditions of uncertainty.

Psychology has already identified many of these functions with remarkable precision. Systems theory may provide the complementary structural language required to represent their interactions, while artificial intelligence may eventually supply the computational framework capable of reasoning over such models. Together, these disciplines may offer a new methodological approach to understanding human behaviour without reducing its inherent complexity.

Whether a comprehensive systems model of human behaviour can ultimately be developed remains an open question. The purpose of this article has not been to provide definitive answers, but to propose a direction for future interdisciplinary research. If language models transformed the way machines process human communication, systems models may one day transform the way artificial intelligence understands the human systems behind that communication.

The question is therefore no longer whether artificial intelligence can analyse conversations. It already can. The more fundamental question is whether we are ready to model the human systems from which those conversations emerge.

---

About The Author

Ralph Larson is the authorial identity behind a collection of essays exploring sustainable computing, Linux, digital autonomy and responsible artificial intelligence. His work is published through the research initiatives EBAN Research Lab, DDR Lab and Independent Edition.

---